01 the business
Family-owned machine shop
65 staff, two senior machinists near retirement
Two machinists with thirty years' experience know which machine needs a trick and which customer wants things done a certain way. They plan to retire in the next two years. Nobody has had time to write anything down.
02 before
The problem
In many small firms, the hardest knowledge lives in one or two people's heads: which machine needs a trick, which client wants things done a certain way. When they retire or leave, it goes with them. Nobody has time to write it down while they are still there.
- Critical know-how lives in two people's heads.
- Juniors interrupt them constantly.
- Nothing is written down.
03 the build
What the AI does
- Step 1: Plans a series of short interviews around the person's role, systems and regular problems.
- Step 2: Asks follow-up questions by chat or voice, and records walkthroughs over Zoom or Teams.
- Step 3: Turns the answers into how-to notes, checklists and 'if this, then that' guides.
- Step 4: Links each note to the related documents, machines, clients or jobs in your file storage.
- Step 5: Answers colleagues' questions from these notes in Slack or Teams and cites the source.
paused: waiting for a person to approve
How it was put together
- Planned short interviews around each person's machines, customers and common problems.
- Asked follow-up questions and recorded walkthroughs on the shop floor.
- Turned answers into how-to notes, checklists and 'if this, then that' guides.
- Linked notes to machines, jobs and documents, and answered colleagues' questions from them.
04 human in the loop
Where a person stays in charge
The person interviewed reviews and corrects every note before it is shared. Their manager decides who can see sensitive client or pricing details.
- The expert reviews and corrects every note before it is shared.
- Managers decide who can see sensitive customer or pricing details.
- Answers cite the note they came from.
05 rollout
From kickoff to live in 2 months of capture
- Weeks 1–2Plan topics with the experts and their manager.
- Weeks 3–8Two short sessions a week and note reviews.
- OngoingOpen the notes to colleagues and keep adding.
06 after
What changes
Hard-won experience stays in the business after people move on. New staff can ask a question and get the answer an old hand would have given.
Assume 20 key topics per expert, two experts, and 3 hours per topic of interviews, review and walkthroughs.
20 × 2 × 3 hours = 120 hours of capture spread over a few months, against decades of experience leaving the firm.
What to measure
- Topics captured and reviewed.
- Questions answered from the notes.
- Interruptions to the experts.
- Time for juniors to run each machine alone.
07 evidence
What the research says
Across 5,179 customer support agents, an AI assistant raised issues resolved per hour by 14% on average and by 34% for novice and lower-skilled agents.
NBER, Generative AI at Work (Brynjolfsson, Li and Raymond) (opens in a new tab) · 2023
Knowledge workers spend roughly a fifth of their working week searching for and gathering information.
McKinsey Global Institute, Capturing business value with social technologies (opens in a new tab) · 2012
08 lessons
Mistakes to avoid
- Long sessions. Short and frequent works better.
- Skipping expert review.
- Starting too late. Begin well before the leaving date.
09 questions
Common questions
- Will experts feel replaced?
- Usually the opposite. Their knowledge is recorded with their name on it, and they get fewer interruptions.
- Does research support AI help for less experienced staff?
- Yes. A study of 5,179 support agents found AI help raised productivity by 34% for novice workers, by spreading what experienced colleagues knew.